Chasing Protons: How Isothermal Titration Calorimetry, Mutagenesis

May 26, 2014 - Drug molecules should remain uncharged while traveling through the body and crossing membranes and should only adopt charged state upon...
0 downloads 11 Views 3MB Size
Article pubs.acs.org/jmc

Chasing Protons: How Isothermal Titration Calorimetry, Mutagenesis, and pKa Calculations Trace the Locus of Charge in Ligand Binding to a tRNA-Binding Enzyme Manuel Neeb,† Paul Czodrowski,‡ Andreas Heine,† Luzi Jakob Barandun,§ Christoph Hohn,§ François Diederich,§ and Gerhard Klebe*,† †

Institut für Pharmazeutische Chemie, Philipps-Universität Marburg, Marbacher Weg 6, 35032 Marburg, Germany Computational Chemistry, Merck Discovery Technologies, Frankfurter Straße 250, 64293 Darmstadt, Germany § Laboratorium für Organische Chemie, ETH Zurich, Vladimir-Prelog-Weg 3, 8093 Zurich, Switzerland ‡

S Supporting Information *

ABSTRACT: Drug molecules should remain uncharged while traveling through the body and crossing membranes and should only adopt charged state upon protein binding, particularly if charge-assisted interactions can be established in deeply buried binding pockets. Such strategy requires careful pKa design and methods to elucidate whether and where protonation-state changes occur. We investigated the protonation inventory in a series of linbenzoguanines binding to tRNA−guanine transglycosylase, showing pronounced buffer dependency during ITC measurements. Chemical modifications of the parent scaffold along with ITC measurements, pKa calculations, and site-directed mutagenesis allow elucidating the protonation site. The parent scaffold exhibits two guanidine-type portions, both likely candidates for proton uptake. Even mutually compensating effects resulting from proton release of the protein and simultaneous uptake by the ligand can be excluded. Two adjacent aspartates induce a strong pKa shift at the ligand site, resulting in protonation-state transition. Furthermore, an array of two parallel H-bonds avoiding secondary repulsive effects contributes to the high-affinity binding of the lin-benzoguanines.



INTRODUCTION In recent time, isothermal titration calorimetry (ITC) has become a powerful tool in life sciences with a broad scope of application, e.g., to obtain insight into the energetics or kinetics of binding, permeation through lipids as well as the characterization of low or high affinity binding ligands that are not accessible by direct titrations.1−4 In structure-based drug design, this method is of utmost importance to obtain not only information about binding affinities but also to access the entire thermodynamic signature of a ligand within one experiment.5 In addition, it gives valuable insight into changes of protonation states upon protein−ligand binding when the measurements are carried out in different buffer systems showing deviating contributions to the heat of ionization ΔHion.6 Numerous studies have underlined the importance to investigate the protonation state of bound ligands to their target enzyme.7−13 In fact, the ITC method quantifies the molar amount of protons transferred from or released to the buffer environment. Unfortunately, it does not provide any information about the exact protonation site. Supporting the ITC measurements by pKa calculations and mutational studies can help to detect the residue or ligand functional group responsible for the protonation reaction. pKa calculations © XXXX American Chemical Society

suggest the likely candidate residues of interest and subsequent titrations of the specifically generated mutants can verify the protonation sites experimentally.9,10 In an unfortunate situation, no overall change in protonation can be observed, even though different buffers have been used and protons are transferred. Recently, Baum et al. observed the simultaneous superposition of a proton uptake and release reaction leading overall to a balanced proton inventory during ITC titrations.8 This study underlines the importance to investigate congeneric ligand series, which differ only by small gradual changes varying the basic or acidic properties, respectively. On the one hand, understanding the properties of titratable groups that can change their protonation state upon binding, both on protein and ligand sites, is an important prerequisite to correctly assign and subsequently interpret the thermodynamic signature of a compound. For instance, the enthalpic signal recorded in a titration comprises the heat contributions of all interactions that are formed between the ligand and the amino acids of the target protein as well as those to trapped or released water molecules or ions. Depending on the type of interaction ranging from simple hydrogen bonds between Received: March 14, 2014

A

dx.doi.org/10.1021/jm500401x | J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry

Article

our initial measurements (Scheme 1). This compound exhibits several basic functional groups appropriate to pick-up a proton,

uncharged species to charge-assisted ones and hydrogen bonds with salt-bridge character, their strength and thus their contribution to the enthalpic signal of the binding event can vary strongly. Taking continuum electrostatics as a reference, this fact can be easily understood. If charge-assisted contacts are formed in an environment of low local dielectric conditions, which are usually given in deeply buried binding pocket of significant hydrophobic character, they will exhibit large contributions. This results from the fact that electrostatic forces, described by Coulomb’s law, show the product of the charges in the numerator and, apart for the squared contact distance (if the potential is considered, it is inverse-linear in distance), the dielectricity constant ε in the denominator. Thus, high dielectricity as given in an aqueous environment (ε ∼ 80) attenuates charge-assisted contacts strongly. In hydrophobic environment, where ε is small (ε ∼ 1−4), an increasingly exothermic signal is experienced as there the involved interacting groups are shielded from the solvent environment. The thermodynamic signature of protein−ligand binding can become quite complex. For the interpretation of such data, it is therefore of utmost importance to know whether and where possible changes in protonation states might create chargeassisted contacts. If these contacts are planned and subsequently established in buried binding pockets, they can contribute significantly to the exothermic binding signature and thus enhance affinity in a tailored fashion. On the other hand, proper adjustment of protonation states plays an important role for the permeation and distribution of the potential drug molecule to obtain access to a target cell under consideration. It is well-known that permeation depends on multiple factors such as lipophilicity, hydrogen-bonding properties, which are important for the transfer from the water phase into the lipid bilayer, or the shape of a drug molecule, and last but not least its charge.14−18 Drug molecules often contain functional groups of acidic and basic character and thus occur dependent on their pKa values and the local pH conditions of the surrounding compartment as charged or neutral species in equilibrium. The uncharged form of a drug will permeate to a greater extent through biological membranes than its corresponding charged form. Accordingly, it would be highly desirable that a drug molecule remains uncharged as long as it is transported via membranes and adopts a charged state only upon binding to the target protein as usually higher affinity results when charged species interact with one another (see above). In the present study we investigated a series of ligands comprising a 6-amino-1,7-dihydro-8H-imidazo[4,5-g]quinazolin-8-one (lin-benzoguanine) and a 1,7-dihydro-8Himidazo[4,5-g]quinazolin-8-one (lin-benzohypoxanthine) scaffold with respect to their inhibitory potency of the tRNA modifying enzyme tRNA−guanine transglycosylase (TGT) from Zymomonas mobiliz.19,20 TGT has been shown to be a putative target to fight the pathogenicity of shigellosis, a severe diarrheal disease.21,22 Previous comparisons of the linbenzoguanines and lin-benzohypoxanthines already showed surprising differences concerning their binding modes and their binding affinities,20 which prompted us to take a closer look at the features of the two scaffolds in the TGT-bound state.

Scheme 1. Chemical Formulae of the Investigated Ligandsa

a

lin-Benzoguanine 1 changes protonation state upon binding and exhibits three potential protonation sites (colored circles). Each of these sites is altered and possibly removed by chemical means in the ligands 2−4.

namely N(5) of the aminopyrimidinone substructure (encircled in orange), N(3) of the attached aminoimidazole ring (encircled in green), and the nitrogen of the morpholino substituent (encircled in blue). A crystal structure of TGT in complex with 1 was determined to a resolution of 1.42 Å (Figure 1A). The linbenzoguanine scaffold is well-defined in the |Fo| − |Fc| difference electron density, and a π-stacking interaction between the side chains of Tyr106 and Met260 can be observed. Polar interactions of the aminopyrimidinone ring are experienced with the side chains of Asp102, Asp156, Gln203, and the backbone NH of Gly230 as analogously found for other members of this compound series.19 The guanidinium moiety of the aminoimidazole ring forms interactions to the closely adjacent backbone carbonyl oxygens of Ala232 and Leu231 and contacts a conserved water molecule, which itself is part of a crucial cluster of water molecules.23 The two guanidine-type portions of the parent purine scaffold are highly buried in the protein and thus fully shielded from solvent access (aminopyrimidinone portion, 100% buried; aminoimidazole portion, ∼96% buried, if the conserved water molecule is excluded from the calculation). We observed a reasonably well-defined electron density for the entire ligand. Although the average B-factor of the morpholino moiety is approximately three times larger than that of the tricyclic scaffold (Bmorpholine = 21.5 Å2 vs Btricycle = 9.1 Å2), the N+-CH2-CH2-NH linker to the morpholine ring is observed in a gauche-conformation partly occupying the flat ribose-33 subpocket and forming a weak charge-assisted hydrogen bond (O···N distance: 3.3 Å, also in the following H-bond contacts refer to the distance between the involved non-hydrogen atoms) via its morpholino oxygen to the guanidinium group of Arg286. The morpholine ring adopts a chair conformation and was refined in the shown orientation



RESULTS AND DISCUSSION Selected Ligands and Binding Mode of lin-Benzopurines to Z. mobilis TGT. To study putative changes in the protonation states of lin-benzopurines, 1 has been selected for B

dx.doi.org/10.1021/jm500401x | J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry

Article

Figure 1. Binding modes of TGT·1 (PDB ID: 4PUJ) and TGT·2 (PDB ID: 4PUK). The protein is represented as cartoon. The ligand and interacting residues are represented as sticks (carbon Protein = gray, carbon Ligand = blue/green, nitrogen = blue, oxygen = red). For clarity, the πstacking residues Tyr106 and Met260, which flank the tricyclic core, are not shown. The lin-benzoguanine scaffold located in the guanine-34 binding pocket is well-defined in the |Fo| − |Fc| difference electron density (green) at a σ level of 2.5. The ligands form several interactions to Asp102, Asp156, Gln203, Gly230, Leu231, and Ala232 (dashed lines). The hydrogen bond to Leu231 is enabled by a ligand induced backbone-flip, which is stabilized by Glu235. (A) The morpholinoethyl substituent shows residual mobility in the bound state indicated by the less well-defined electron density. Nevertheless, it was refined to an occupancy of 80%. The linker to the morpholine substituent adopts a gauche-conformation and forms a weak charge-assisted hydrogen bond of 3.3 Å between the oxygen of the morpholine ring and the guanidinium group of Arg286. To minor extent, the residual density indicates the fully extended anti-conformer. (B) Due to the lack of the extended 2-substituent, the methyl group interacts with the solvent (green spheres) showing C···O distances of up to 4.1 Å.

Protonation States of lin-Benzopurine·TGT Complexes. To obtain insights into the potentially superimposed protonation effects that occur upon ligand binding, ITC measurements were performed using Hepes, Tris, and tricine buffer at pH 7.8. The three buffers differ in their enthalpy of ionization.27,28 The observed enthalpies for the complexation of the studied ligands were plotted against the heat of ionization of the buffers (Figure 2), and the recorded thermodynamic profiles are listed in Table 1. Ligand 1 picks up 0.95 protons per

with an occupancy of 80%. To a minor extent the residual density indicates the fully extended substituent with an anticonformation of the linker. Since a broad series of lin-benzoguanines has been synthesized during our studies,19,20,23,24 we considered selected ligands in the present work, either avoiding or decreasing the basic character of the scaffold and the substituents under consideration (Scheme 1). In ligand 2, the basic morpholinoethyl substituent is replaced by a simple methyl group. The binary complex with Z. mobilis TGT was obtained with a maximum resolution of 1.49 Å (Figure 1B). The scaffold adopts exactly the same binding mode forming similar interactions to 1. Due to the lack of the extended 2-substituent, the methyl group is in direct contact with five water molecules showing C···O distances up to 4.1 Å. In addition to the changes found in lin-benzoguanine 2, the exocyclic amino function at the pyrimidinone ring has been removed in lin-benzohypoxanthine 3. Barandun et al. recently reported on the binding mode of this class of linbenzopurines.20 In contrast to binary complexes observed with the lin-benzoguanines, Asp102 adopts a downward-rotated orientation pointing away from the ligand and the binding pocket. The observed rotamer undergoes hydrogen bonds to the side chain of Asn70 and the backbone NH of Thr71. A similar geometry is found in apo TGT.25 The created empty space between ligand and protein is filled by a cluster of six water molecules. All other interactions are established as similarly found in the TGT complexes with lin-benzoguanines. The strongly basic character of the aminoimidazole moiety in 2-amino-lin-benzopurines 1−3 has been changed to a less basic amidinium portion in 4. Except for the hydrogen bond to the backbone oxygen of Ala232, this compound is able to undergo the same hydrogen bonding pattern as described for the abovementioned crystal structure (PDB ID: 3C2Y).26

Figure 2. ΔHobs is plotted against ΔHion for ligands 1−4 binding to TGT. The icons indicate the mean of the measured enthalpy values for each buffer system in kJ/mol. A linear fit was applied to the data points illustrated as solid lines. The positive slope shows a buffer dependency of 1, 2, and 4, suggesting an uptake of approximately one proton per mole by the complex. Only binding of 3 does not experience any changes in protonation state. The corrected value for ΔHbind can be extracted from the intercept; the number of captured protons results from the slope, which was calculated separately taking the standard deviations of each data point into account. C

dx.doi.org/10.1021/jm500401x | J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry

Article

Table 1. Thermodynamic Profiles of the Investigated Ligands in the Applied Buffer Systems ligand

TGT variant

Kd (nM)

ΔG0 (kJ·mol

−1

)

buffer

35.0 ± 6.9

−42.6 ± 0.5

1275.6 ± 297.3

−33.8 ± 0.6

Hepes tricine Tris

52.4 ± 6.9

−41.6 ± 0.3

Hepes tricine Tris

Asp102Asn

146.1 ± 25.6

−39.0 ± 0.4

Hepes tricine Tris

2

Asp156Asn

188.1 ± 59.1

−38.5 ± 0.8

Hepes tricine Tris

3

wild type

411.0 ± 63.8

−36.5 ± 0.4

Hepes tricine Tris

4

wild type

287.5 ± 52.9

−37.4 ± 0.5

Hepes tricine Tris

1

wild type

1

Glu235Gln

2

wild type

2

c

Hepes tricine Tris

ΔHobs (kJ·mol−1)a −78.3 −65.8 −51.9 −96.4 −51.0 −39.9 −28.6 −66.2 −74.8 −66.7 −50.9 −97.4 −33.5 −35.0 −33.9 −33.9 −51.4 −52.6 −49.2 −53.2 −48.0 −47.7 −47.6 −48.4 −69.3 −57.1 −43.4 −89.2

± 1.2 ± 0.3 ± 1.1 ± 4.0d ± 0.7 ± 0.5 ± 0.3 ± 3.2 ± 1.8 ± 0.1 ± 0.9 ± 2.2 ± 0.4 ± 0.4 ± 0.3 ± 2.1 ± 0.2 ± 2.5 ± 0.4 ± 0.4 ± 0.2 ± 0.2 ± 0.3 ± 0.3 ± 0.8 ± 1.6 ± 0.1 ± 1.4

−TΔS0 (kJ·mol−1)b 35.7 23.2 9.3 53.8 17.9 6.0 −5.7 32.4 33.2 25.1 9.3 55.8 −6.1 −4.0 −4.8 −4.9 13.0 13.4 11.4 12.6 11.5 11.2 11.1 11.9 31.9 19.7 6.0 51.8

± 1.3c ± 0.6 ± 1.2 ± 4.0d ± 0.7 ± 0.5 ± 0.5 ± 3.3 ± 1.9 ± 0.3 ± 0.9 ± 2.2 ± 0.5 ± 0.5 ± 0.4 ± 2.1 ± 0.4 ± 2.5 ± 0.6 ± 0.9 ± 0.4 ± 0.5 ± 0.5 ± 0.5 ± 0.9 ± 1.6 ± 0.5 ± 1.5

The corresponding plot to obtain net ΔHbind of wild type TGT corrected for ionization effects is shown in Figure 2. b−TΔS0 has been calculated according to the Gibbs−Helmholtz equation. cThe errors were estimated by means of standard deviation for Kd and ΔG0 comprising at least six measurements and for ΔHobs at least two. The error for −TΔS0 was calculated according to error propagation. dBuffer corrected data are displayed in bold. a

benzoguanines, a value of 4.4 has been assigned to N(5)H+ and 5.7 to N(3)H+ in lin-benzoguanine 2.20 Remarkably, the acidity of N(5)H+ is shifted by about two logarithmic units to pKa = 1.8 in lin-benzohypoxanthine 3. To confirm the experimentally derived evidence, Poisson− Boltzmann calculations were consulted to analyze shifts of pKa values of the active site residues and the ligand functional groups upon binding.30 Within the 12 Å sphere centered around Cγ of Tyr106, all titratable residues were selected and the corresponding pKa values were calculated prior and after complex formation. The calculations reveal only significant shifts in basic or acidic properties for Asp102, Asp156, and N(3) or N(5) of ligand 1 (Supporting Information Table S2). Accordingly, as a first working hypothesis, either N(3), embedded into an aminoimidazole substructure, or N(5), being part of an aminopyrimidinone moiety, could be likely candidates for the proton uptake. The shift suggested for N(5) appeared larger (ΔpKa = 2.6) compared to N(3) (ΔpKa = 1.0), however, as N(3) shows the more basic character in aqueous solution, both sites appear appropriate to pick up a charge. A previous study on the pKa properties of TGT·2-amino-linbenzoguanine inhibitor complexes was based on the hypothesis that the ligands bind with N(3) in protonated state.24,26 Accordingly, the 2-aminoimidazole moiety of the ligand scaffold would be positively charged and form charge-assisted hydrogen bonds to the backbone carbonyl oxygens of Leu231 and Ala232. This interaction would induce a backbone flip, which is

mole formed complex, which either protonate the protein or the bound ligand (Figure 2, blue). Apart from the basic functionalities of 1, there are several amino acid residues that could also be involved in the proton uptake. Within a 12 Å radius centered at Cγ of Tyr106 (Figure 3) around the binding pocket, Asp102, Asp156, Glu157, Glu235, and Asp280 are putative proton-acceptor groups. According to Poisson−Boltzmann calculations, these residues can be assumed as fully deprotonated at the applied pH conditions between 5.5 and 8.5.26 Asp102 and Asp156 are directly involved in ligand binding (Figure 1). We decided to investigate first the ligand series exhibiting different basic characters. lin-Benzoguanines 2 (0.96 protons per mole; Figure 2, red) and 4 (0.95 protons per mole; Figure 2, green) also suggest uptake of approximately one proton. Only lin-benzohypoxanthine 3 (0.02 protons per mole; Figure 2, purple) is obviously not capable of share these properties. Our findings agree with the experimentally determined pKa values reported for the lin-benzohypoxanthines and linbenzoguanines in aqueous solution (Supporting Information Table S1).20,24 Unfortunately, no values are available for the morpholinoethyl substituent in 1; we therefore assume a value similar to that of parent morpholinoethyl. The value for the latter has been reported to be pKa = 7.7.29 We suppose that this assumption is justified as the morpholino portion is attached via an aliphatic ethyl linker which largely avoids transmission of electronic effects across molecular portions. For 2-amino-linD

dx.doi.org/10.1021/jm500401x | J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry

Article

obviously shifts its pKa properties and induces an alternative conformation of Asp102 to interact with the ligand’s guanidinium group. To assess whether N(3) or N(5) become protonated upon protein binding or whether an even more complex situation with several superimposed and mutually compensating effects is given, we embarked onto a systematic mutational study combined with subsequent ITC experiments and pK a calculations. Mutational Studies. According to our pKa calculations, Asp102 and Asp156 form charge-assisted interactions with the aminopyrimidinone portion of the lin-benzoguanine-type inhibitors and might be involved in the protonation of N(5). Therefore, we decided to replace Asp by Asn at both sites to investigate the impact of these residues on ligand binding. The replacement by Asn avoids the negative charge in this region and introduces a permanently uncharged contact. While the Asp156Asn mutant still exhibits some, however by a factor of 3 significantly reduced catalytic activity compared to the wild type, the Asp102Asn mutant shows no activity at all.31 For both mutants, the crystal structures with 2 were determined and ITC measurements carried out in the three buffer systems. Structural data were recorded to a resolution of 1.65 Å and 1.85 Å. Ligand 2 adopts in the Asp156Asn mutant a similar binding mode to the wild type (Figure 3A,B). The aminopyrimidinone ring interacts with the side chains of Asp102, Asn156, Gln203, and the backbone NH of Gly230. The orientation of the replaced Asn156 residue is slightly shifted compared to the wild type, moving the terminal carboxamide group approximately 16° out of the plane through the linbenzoguanine scaffold. The interaction between the terminal carboxamide oxygen of Asn156 and the ligand’s exocyclic amino function remains unchanged compared to the wild type (Δd = 0.1 Å). However, the adjacent contact via the NH2 of the carboxamide group of Asn156 and N(7) of 2 is extended from 2.7 to 2.9 Å. This expansion results both from the larger radius of the nitrogen atom and the replacement of a charge-assisted by a neutral hydrogen bond. Additionally, the side chain oxygen of residue Gln203 tilts about 13° to the front, elongating the distance between its amino function and the same group in Asn156 from 2.8 to 3.2 Å. The strength of the interaction formed to the carbonyl oxygen of the complexed ligand and Gln203 remains unaffected as the uncharged contact distance suggests. Also, the interactions formed between the ligand and the side chain of Asp102 do not differ from the wild type. To establish this interaction pattern, the proton at N(7) must be transferred to N(5) to produce the other tautomer. In summary, neutral hydrogen bonds are formed to Asn156 and charge-assisted ones to Asp102 (Scheme 2). In the complex structure of the Asp102Asn mutant with 2, the ligand adopts a position similar to the one in the wild type. It is slightly shifted off from the position of Asp156 (Figure 3A). The mutated residue Asn102 does not form hydrogen bonds to the aminopyrimidinone moiety of the ligand but adopts an orientation similar to those of wild-type enzymes with the lin-benzohypoxanthine-type inhibitors. A hydrogen bond is formed to a nearby water molecule (2.8 Å) and the backbone NH of Thr71 (2.7 Å).20 As expected, the binding affinity of 2 determined by ITC measurements drops for both mutant variants as favorable interactions are lost in both cases. For the Asp156Asn mutant, a reduction by a factor of nearly four (wild type, 49 ± 5 nM;

Figure 3. Protein is displayed as cartoon with its interacting residues shown in line representation. The mutated side chains and 2 are highlighted by sticks (nitrogen = blue, oxygen = red). Dashed lines represent hydrogen bonds. Distances are given in Å. (A) Superimposition of 2 as bound to wild-type TGT (cyan; PDB ID: 4PUK), mutant Asp102Asn (yellow; PDB ID: 4PUL) and mutant Asp156Asn (orange; PDB ID: 4PUM). The ligands adopt overall the same orientation in all displayed complexes. The side chain of Asn102 adopts a conformation rotated away from the binding site compared to Asp102 in the wild-type enzyme. To the contrary, Asn156 remains in a similar position as observed for Asp156. (B) Close-up of the interactions of the mutated residue Asn156 (orange) compared to those of wild-type enzyme (cyan). Asn156 is slightly rotated by 16°, extending the interaction to the amino group of Gln203 and N(7) of the ligand. All other connections remain unaffected. (C) Close-up of the interactions of the mutated residue Asn102 (yellow) compared to those of wild-type enzyme (cyan). Asn102 shows the rotated orientation adopting a similar conformation as found for the linbenzohypoxanthine complexes. Instead of interacting similarly with the ligand as Asp102 in the wild type, Asn102 forms contacts to the backbone NH of Asn70 and a nearby water molecule.

stabilized by a further charge-assisted hydrogen bond formed between Glu235 and the backbone NH of Leu231.24,26 The nearly 30-fold increase in binding affinity suggested by a biochemical assay between 2 and 4 was considered as clear indication of this protonation event at the imidazole moiety of 2. However, recent results also support the hypothesis that N(5) is the likely candidate to pick-up a proton. A newly published crystal structure of Z. mobilis TGT in complex with 2-amino-lin-benzohypoxanthine-type inhibitors shows that Asp102 adopts a significantly different orientation in the complexes compared to the corresponding lin-benzoguanine complexes.20 This residue is rotated away from the ligand toward the ribose-34 subpocket as similarly found in the apo enyzme.20,25 In the new position, Asp102 forms an interaction with the carboxamide group of Asn70 and the backbone NH of Thr71.20 By capturing a proton, the lin-benzoguanine moiety E

dx.doi.org/10.1021/jm500401x | J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry

Article

picked-up charge contributes significantly to the enhanced binding affinity? Also, a more complex situation with respect to changes of protonation states might be given in this part of the structure. In this area, Glu235 takes an important role on ligand-binding and protein function. It has been shown that Glu235 adapts its protonation state depending on the orientation of the peptide bond between Leu231 and Ala232 (Scheme 3A,B).25 This peptide bond is exposed to the binding pocket and experiences a backbone flip upon ligand binding or upon changes in the environmental pH value. If a linbenzopurine binds to the active site, the peptide bond’s carbonyl oxygen of Ala232 is oriented into the binding pocket and accepts a hydrogen bond from N(1)H of the ligand (Scheme 3C). In consequence, Glu235 becomes deprotonated in order to accept a hydrogen bond of the flipped backbone NH group.32 Glu235 is the only proximal amino acid involved in ligand binding, which could be able to release a proton. On the basis of these considerations, we mutated Glu235 to Gln to experimentally assess the remaining overall protonation inventory. ITC experiments were carried out with the Gln235 variant in the three mentioned buffer systems, indicating an overall proton uptake of 0.79 protons per mole. This nearly identical protonation inventory speaks for N(5) in the aminopyrimidinone moiety as sole proton acceptor of the linbenzoguanine scaffold. Nonetheless, these experiments do not completely rule out a possibly more complex situation. Glu235 has to release a proton in consequence of the peptide bond flip. This only makes this bond able to recognize the linbenzopurine scaffold. It could be imagined that the proton release of Glu235 is combined by a simultaneous protonation of N(3) in the aminoimidazole portion. In this case, both effects would mutually compensate and nullify in the inventory. The following experiment makes this consideration unlikely. Previously, we were able to determine the apo structures of wild type TGT at pH 5.5 and 8.5.25,33 At pH 5.5, the NH group of the Leu231/Ala232 peptide bond is oriented toward the binding pocket and Glu235 must adopt a protonated state to stabilize the peptide bond (Scheme 3A). At pH 8.5, the flipped peptide bond orientation is given and Glu235 has to adopt a deprotonated state (Scheme 3B). We therefore took crystals grown at pH 5.5 and transferred them in a gradual stepwise fashion to higher pH, finally reaching a value of 7.8. Subsequently, we subjected the thus treated crystal to a structure determination and observed the geometry of the peptide bond with the backbone carbonyl group oriented toward the binding pocket (Scheme 3B). This geometry is only compatible with Glu235 present in the deprotonated state. We thus conclude that the backbone flip must occur already at lower pH and ligand binding studied by our ITC titrations occurs to the enzyme with the peptide bond in the orientation for binding the ligand and with Glu235 in the deprotonated state. Therefore, a simultaneous compensating protonation change of N(3) and Glu235 can be excluded. Finally, there remains the question why the addition of the exocyclic C(2)−NH group at the imidazole moiety of the linbenzopurines (2 compared to 4) results in a significant affinity enhancement, which made Ritschel et al. believe that this part of the ligand would bear a charge improving the affinity contribution of the formed additional hydrogen bonds.26 The biochemical assay exaggerated this improvement (factor 30) as the ITC experiments suggest a smaller value (factor 5.5). Nonetheless, the enhancement is still remarkable. It has to be regarded that both NH functionalities of the 2-aminoimidazole

Scheme 2. Interactions Formed between 2 and Amino Acids 102 and 156 of Wild-Type TGT (Cyan), the Asp156Asn Mutant (Orange), and the Asp102Asn Mutant (Yellow)a

a

In the TGT·2 complex, N(5) of the parent scaffold becomes protonated upon binding and forms salt bridges to Asp102 and Asp156. To bind to the Asp156Asn mutant in a similar way, the proton at N(7) must be transferred to N(5), then forming hydrogen bonds to Asn156 and charge-assisted ones to Asp102. In the Asp102Asn·2 complex, Asn102 is rotated off the binding pocket. Therefore, N(7) carries the proton to form charge-assisted hydrogen bonds to Asp156.

mutant, 188 ± 59 nM) is found, whereas a 3-fold loss is recorded for the Asp102Asn mutant (146 ± 26 nM). Surprisingly, 2 does not show any significant buffer dependency upon binding to the two mutants (Table 1). Also, Poisson−Boltzmann calculations performed on these mutant complexes support this hypothesis (Supporting Information Table S3−S4). The close contacts to the two negatively charged aspartates in the wild type reinforce protonation of the ligand and cause the strong pKa shift. Remarkably, in the Asp102Asn mutant, the presence of Asp156 as charged residue and the neutral Asn102, which is rotated away from the binding site is not sufficient to induce an equally strong pKa shift as experienced in the wild type. This finding also correlates with the expanded interactions of Asn156 and the ligand’s aminopyrimidinone moiety in the Asp156Asn mutant where no positive charge is located on this part of the ligand. Up to this point all experimental findings indicate a protonation of N(5) in the aminopyrimidinone moiety. Does this, however, rule out the former hypothesis that N(3) in the aminoimidazole portion changes protonation and that the F

dx.doi.org/10.1021/jm500401x | J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry

Article

Scheme 3. Dependence of the Peptide Bond between Leu231 and Ala232 on the pH Value and Ligand Bindinga

a TGT apo structures of crystal grown at different pH values clearly indicate that a backbone flip between Leu231 and Ala232 occurs dependent on the environmental pH value and the ligand bound. (A) Scheme of the backbone orientation at pH 5.5. The backbone NH group of Leu231 is facing the binding pocket. The carboxyl group of Glu235 is present in its protonated state interacting with the carbonyl function of Leu231 and the backbone NH group of Val233. (B) Scheme of the backbone orientation at pH 8.5. The backbone carbonyl function of Leu231 is exposed to the binding pocket. Hence, the side chain of Glu235 is accepting two hydrogen bonds from the backbone NH groups of Ala232 and Val233 and must occur in its deprotonated form. The crystal structure determined at pH 7.8 and obtained by a gentle pH transition from pH 5.5 shows the same backbone orientation as at pH 8.5. This clearly indicates that Glu235 is already deprotonated in the apo structure at pH 7.8. (C) Scheme of the ligand bound backbone orientation. Taking the orientation of the peptide bond between Leu231 and Ala232 in the apo structure at pH 7.8 into account, Glu235 accommodates a bound ligand without release of a proton at this pH value. (D) Overlay of TGT apo structures at pH 5.5 (PDB ID 1P0D), 7.8 (PDB ID 4PUN), and 8.5 (PDB ID: 1PUD). Amino acid residues are shown in stick representation (nitrogen = blue, oxygen = red). For clarity, the side chain of Val233 is not shown.



CONCLUSION In the present study, changes of protonation states of ligands binding to the active site of the enzyme TGT were investigated. ITC measurements supported by pKa calculations confirmed the uptake of one proton in case of the lin-benzoguanine-type ligands. The structurally related lin-benzohypoxanthines do not show such a pick-up of a proton. Considering the adopted binding modes, the lin-benzohypoxanthines do not induce a relocation of the side chain of Asp102, which binds in case of the lin-benzoguanines directly to the ligand via two short hydrogen bonds. This relocation of Asp102 and the presence of Asp156 in short distance to the aminopyrimidinone moiety of the ligand creates a negatively charged environment. This provokes a significant pKa shift of the ligand’s heterocyclic moiety, which therefore becomes protonated and thus positively charged. Mutational exchanges of both Asp residues by Asn underline that the cluster of negative charges is responsible for this induced pKa shift. However, the intrinsic pKa values of the uncomplexed ligand under consideration must fall into a crucial window to achieve such a change in protonation upon binding. Both the linbenzoguanines and the lin-benzohypoxanthines exhibit a second guanidinium type motif in the five-membered imidazole portion. Our investigations reveal, however, that N(3) of the aminoimidazole moiety does not change protonation state upon binding but the moiety experiences favorable hydrogen

portion forming the hydrogen bonds to the neighboring carbonyl groups of Leu231 and Ala232 are mutually adjacent (Scheme 3C). This dual hydrogen bonding array involves a pattern of two donor and acceptor groups with parallel orientation. Such an arrangement avoids secondary repulsive interactions that, for example, are given in the above-discussed contact of the aminopyrimidinone moiety and Asn156 (Scheme 2, center). The separation distance of the hydrogen atoms in the directly adjacent hydrogen bonds is rather short, and thus substantial electrostatic interactions will occur.34 Partial positive charges are given on the hydrogen atoms, whereas partial negative charges are experienced on the nitrogen and oxygen atoms in the NH···O hydrogen bonds. A favorable situation is given, as in the present case, where one of the binding partners bears all the hydrogen donor groups, the other all acceptor groups. The situation is worse, if donor and acceptor sites alternate between both partners, because then secondary repulsive interactions results from complexation. This incident is given in the contact to Asn156 and clearly responded by the structure with expanded and distorted geometry (Figure 3B). The enhancement of adjacent hydrogen bonding contacts following the DD/AA pattern has already been described in host−guest complexes many years ago.35 G

dx.doi.org/10.1021/jm500401x | J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry

Article

Table 2. Oligonucleotides Used in DNA Mutagenesis

a

oligoname

sequence (5′ to 3′)a

Asp156Asn_f Asp156Asn_b

5′-ATT GTT ATG GCA TTT AAT GAA TGT ACC CCG TAT-3′ 5′-ATA CGG GGT ACA TTC ATT AAA TGC CAT AAC AAT-3′

The mutated codon is underlined. μM for titrations of lin-benzohypoxanthines, each containing 3% DMSO. Due to their low solubility, the ligands were first dissolved in 100% DMSO and diluted with the buffer solution to a final DMSO concentration of 3%. The ligand concentration in the syringe was adjusted to 200−300 μM with the experimental buffer. To examine a potential buffer dependency of the binding process, experiments were carried out in three different buffer systems containing either 50 mM Hepes, Tris or tricine, 200 mM NaCl, and 0.037% Tween 20, pH 7.8 at least in duplicates. The experimentally determined enthalpy ΔHobs was plotted against the heat of ionization ΔHion as reported in literature and fitted by linear regression.28,29 The intercept of the y axis represents the enthalpy corrected for buffer contributions, and the slope discloses the number of protons released or taken up per mole formed complex upon binding according to the equation ΔHobs = ΔHbind + nH+ΔHion. A positive slope indicates uptake of protons by the protein and/or ligand, and a negative slope suggests proton release. All ITC experiments were run at 25 °C after a stable baseline had been achieved. The reference cell contained filtered demineralized water. The initial delay before the injections were started, and the spacing between each injection was adjusted to 180 s. The first injection contained 0.3−0.5 μL of the ligand solution followed by 14− 24 injections of 1.0−2.0 μL. A stirring speed of 1000 rpm was chosen. Raw data were collected as released heat per time. To analyze the raw data using the Origin 7.0 software, the baseline and integration limits were adjusted manually. After integrating the area under the peaks, the first data point was removed due to a reduced accuracy.37 The influence of the heat of dilution was corrected by considering heat contributions collected after saturation of the protein. Kd (dissociation constant) as well as ΔH0 (enthalpy of binding) were extracted by applying a single-site binding model as provided by the manufacturer. Subsequently, −TΔS0 has been calculated according to the Gibbs−Helmholtz equation. Synthesis. Ligands were synthesized and purified as described in detail elsewhere.20,24 Site-Directed Mutagenesis. Site-directed mutagenesis for mutant TGT Asp156Asn was carried out using the QuikChange Lightning kit (Stratagene) following the manufacturer’s instructions. To introduce the mutation into the Z. mobilis TGT expression plasmid pPR-IBA2ZM-WT, the oligonucleotides listed in Table 2 were used. After plasmid preparation using the peqGOLD Plasmid Miniprep Kit II (PEQLAB) according to the vendor’s instructions, the presence of the desired mutation as well as the absence of further mutations were confirmed by sequence analysis (Eurofins MWG Operon, Ebersberg, Germany). Subsequently, the mutated plasmid was transformed into Escherichia coli Rosetta 2(DE3) cells. For the expression of the TGT mutant Asp102Asn, the plasmid pET9d-ZM4-Asp102Asn31 was used after transformation into E. coli BL21(DE3)pLysS cells, for TGT mutant Glu235Gln the plasmid pET9d-ZM4-Glu235Gln32 after transformation into E. coli BL21(DE3)pLysS cells. Protein Expression and Purification. Wild-type TGT as well as TGT mutants Asp102Asn and Glu235Gln were overexpressed and purified as described in detail elsewhere.38,39 The plasmid encoding Z. mobilis TGT mutant Asp156Asn was ligated into the NheI and EcoRV restriction sites of the pPR-IBA2 expression vector (IBA), which encodes a T7 promotor and adds strep-tag II to the N-terminus of the protein. This plasmid was transformed into E. coli Rosetta 2(DE3) cells for protein overexpression. The bacterial cells were grown in two liters 2xYT medium containing 100 mg·L−1 ampicillin and 34 mg·L−1 chloramphenicol at 37 °C and 220 rpm shaking to an optical density at 600 nm (OD600) of 0.8. Isopropyl β-D-1-thiogalactopyranoside (IPTG) was added in a

bond contacts to the protein avoiding secondary repulsive interactions. In summary, the lin-benzoguanines bear no charge in aqueous solution at pH 7.8 but become positively charged at the binding site. The charge is mainly distributed over the guanidine motif of the aminopyrimidinone and only minor effects might influence the aminoimidazole portion. This observation explains the high potency of the linbenzoguanines showing binding affinities down to the one-digit nanomolar range. Their binding is characterized by a strong enthalpic binding signature resulting from the salt bridges formed between the aminopyrimidinone moiety and the deprotonated side chains of Asp102 and Asp156 (Table 1). The contacts to both acidic residues are established in a deeply buried binding pocket that shields the formed charge-assisted interactions from solvent access. In this environment of low local dielectricity, the formed charged contacts experience an enhanced affinity contribution.36 In addition, the hydrogen bonds formed between the aminoimidazole ring and the backbone carbonyl oxygens of Leu231 and Ala232 are expected to be energetically favorable avoiding repulsive effects. The significant loss in binding affinity in case of the linbenzohypoxanthines results from a loss of the interaction to Asp102, which rotates away from the ligand recognition site toward the ribose-34 binding pocket. As a consequence, the pyrimidinone ring remains uncharged, as found in aqueous solution prior to binding. Obviously, Asp156 alone does not take sufficient impact on the adjacent pyrimidinone moiety of the lin-benzohypoxanthines to induce a sufficiently large pKa shift to result in protonation and consequently charge-assisted interactions. With respect to bioavailability, the aminopyrimidinone portion in the lin-benzoguanines exhibits the desirable feature of getting protonated upon protein binding only. Despite this advantage, overall the molecular properties of the ligands are not yet ideal regarding membrane transportation due to a large polar character of the molecules. Both the lin-benzoguanines and the lin-benzohypoxanthines show unsatisfactory membrane permeation in PAMPA measurements.20 This study demonstrates the importance to analyze the protonation properties of structurally closely related ligands. ITC titrations, performed in different buffer conditions, allow elucidating the net protonation inventory and stoichiometry. These studies, however, have to be complemented by sitedirected mutagenesis and pKa calculations to trace where the protons go and accordingly the charges reside on the formed complex. They also help to avoid false conclusions resulting from mutually compensating effects and allow to localize the hot spots of binding and to tailor ligand properties so that transformation to charged state only occurs upon target binding.



EXPERIMENTAL SECTION

Isothermal Titration Calorimetry. ITC measurements were performed using a Microcal iTC200 microcalorimeter system (GE Healthcare). The protein was dissolved in the experimental buffer to a final concentration of 10 μM for titrations of lin-benzoguanines or 20 H

dx.doi.org/10.1021/jm500401x | J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry

Article

Table 3. Data Collection, Processing and Refinement Statistics for the Crystallographically Investigated TGT−Ligand Complexes crystal data (A) Data Collection and Processing collection site no. crystals used λ [Å] space group unit cell parameters a [Å] b [Å] c [Å] β [deg] (B) Diffraction Dataa resolution range [Å] unique reflections R(I)sym [%]b completeness [%] redundancy I/σ (I) Wilson B-factor [Å2] Matthews coefficient [Å3/Da] (C) Refinement resolution range [Å] reflections used for Rfree reflections used for Rwork final R valuesa Rfree [%]c Rwork [%]d no. of atoms (non-hydrogen) protein atoms water molecules ligand atoms RMSD, angle [deg] RMSD, bond [Å] Ramachandran plote most favored regions [%] additionally allowed regions [%] generously allowed regions [%] mean B-factors [Å2] protein atoms water molecules ligand atoms

TGT·1 4PUJ

TGT·2 4PUK

Asp102Asn·2 4PUL

Asp156Asn·2 4PUM

Apo pH 7.8 4PUN

BESSY 14.2 1 0.91841 C2

BESSY 14.2 1 0.91841 C2

PETRA P14 1 1.23953 C2

PETRA P14 1 1.23953 C2

PETRA P14 1 0.97627 C2

90.6 64.8 71.0 96.2

90.8 65.0 71.0 96.3

89.7 64.2 70.4 92.9

89.7 64.7 70.8 93.1

90.5 64.7 69.9 96.0

30−1.42 (1.44−1.42) 76 897 (3785) 5.7 (50.5) 99.3 (98.4) 5.2 (5.1) 28.6 (3.9) 12.8 2.4

30−1.49 (1.52−1.49) 66 573 (3339) 5.1 (48.1) 99.4 (98.8) 5.1 (4.9) 29.3 (3.4) 15.7 2.4

80−1.65 (1.85−1.65) 43 729 (9984) 2.9 (27.5) 91.6 (73.8) 3.1 (2.6) 22.3 (3.8) 22.4 2.4

80−1.93 (1.98−1.93) 30 236 (2238) 4.7 (50.5) 99.0 (99.4) 4.7 (4.6) 21.6 (3.4) 25.8 2.4

50−1.25 (1.30−1.25) 107 127 (11652) 3.0 (50.0) 96.8 (95.3) 3.3 (3.3) 18.0 (2.7) 14.5 2.4

29.5−1.42 2000 74 896

27.3−1.49 2000 64 506

70.3−1.65 1312 42 417

70.7−1.93 1512 28 724

45.0−1.25 5357 101 770

16.0 12.9

17.2 13.9

18.9 16.7

20.2 15.9

16.2 13.9

2911 388 24 1.0 0.007

2922 365 17 1.3 0.014

2829 261 17 1.1 0.008

2854 219 17 1.1 0.009

2920 373

94.2 5.4 0.3

95.3 4.4 0.3

95.9 3.8 0.3

95.9 3.8 0.3

94.0 5.6 0.3

14.3 32.5 13.8

19.4 33.3 16.7

24.6 33.7 22.6

25.1 31.0 24.9

18.5 29.8

1.4 0.013

Values in parentheses are statistics for the highest resolution shell. bR(I)sym = [∑h ∑i|Ii(h) − ⟨I(h)⟩|/∑h ∑i Ii(h)] × 100, in which ⟨I(h)⟩ is the mean of the I(h) observation of reflection h. cRwork = ∑hkl|Fo − Fc|/∑hkl|Fo|. dRfree was calculated as shown for Rwork but on refinement-excluded 3− 5% of data. eStatistics from PROCHECK.50 a

concentration of 1 mM to induce the overproduction of the TGT mutant for 16 h at 15 °C. Cells were harvested by centrifugation (10000 rpm at 4 °C) and resuspended in 50 mL of lysis buffer containing 20 mM Tris, 10 mM EDTA, and 1 mM DTT, pH 7.8. The suspension was sonicated six times for 90 s and subsequently centrifuged for 1 h at 20000 rpm. The clear supernatant was applied onto a Q Sepharose XK26 column (Amersham Biosciences) and treated analogously to the protocol of Romier et al.31 Fractions containing TGT were identified by SDS gel electrophoresis40 and loaded onto a StrepTactin Sepharose column (IBA). After a washing step with a buffer containing 10 mM Tris, 150 mM NaCl, 1 mM EDTA, and 1 mM DTT at pH 7.8, the protein was eluted with 2.5 mM desthiobiotin in the same buffer at a flow rate of 2 mL/min. The sample was dialyzed (cutoff 4000−6000)

to higher salt conditions (500 mM) and the strep-tag II cleaved with the Thrombin Cleavage Capture Kit (Novagen) for 16 h following the manufacturer’s instructions. It was of utmost importance for protein stability to cleave the tag before concentrating the sample using VIVASPIN20 centrifugal concentrators (Sartorius) with a cutoff of 30000 and exchanging the buffer to high salt conditions (10 mM Tris, 2 M NaCl, 1 mM EDTA, pH 7.8). pKa Calculations. A consistent charge model was produced by a modified version of the charge distribution algorithm that initially was developed by Gasteiger and Marsili, named “partial equalization of orbital electronegativities” (PEOE).41 According to previous studies on the target, a dielectric constant of ε = 20 was chosen to describe the properties of the binding pocket considering an implicit solvent model.26 I

dx.doi.org/10.1021/jm500401x | J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry

Article

For the calculation of the pKa values, all titratable groups within a radius of 12 Å around the active site were considered for the site−site interaction portion of the pKa calculations (Cγ of Tyr106 was taken as center of the selection). This identified the following residues as titratable groups: Lys52, Tyr72, His73, Asp/Asn102, Tyr106, Asp/ Asn156, Glu157, Cys158, Tyr161, Glu173, Tyr226, Glu235, Asp238, Glu239, Asp245, Tyr258, Lys264, Asp266, Asp267, Asp280, Cys281, Tyr354, and Tyr381. With the program REDUCE, all hydrogens were added to the protein in which all acidic residues were deprotonated and basic ones were protonated.42 For the uncharged state of the ligand, SYBYL atom types were assigned and an explicit charge of 0 was set for the hydrogen atom of the titratable group prior to the partial charge calculation. After this preparation, the Poisson−Boltzmann calculation was started using the program MEAD.43 The resulting pKa shifts are listed in Supporting Information Table S2−S4 at pH 7.8. Z. mobilis TGT Crystallization and Ligand Soaking. Crystals for soaking were grown using the sitting drop vapor diffusion method at room temperature. The protein solution was adjusted to 12 mg· mL−1 by dilution with high salt buffer (10 mM Tris, 2 M NaCl, 1 mM EDTA, pH 7.8) and mixed with 1.5 μL of reservoir solution (100 mM MES, pH 5.5, 10% (v/v) DMSO, 11−13% (w/v) PEG 8000) to a 3 μL droplet. The reservoir contained 1.0 mL of the above-mentioned solution. Within 1 week, crystals grew to a size appropriate for soaking at 291 K. The selected pregrown crystals were transferred to a 3 μL droplet of reservoir solution mixed with the stock solution containing the desired ligand first dissolved in 100% DMSO and subsequently diluted to a final concentration of 1 mM. The droplet was sealed against 1.0 mL reservoir solution and soaking performed overnight. pH Soaking. For the pH soaking, a single crystal grown at a pH of 5.5 was transferred and incubated over a period of 5 min into droplets consisting of reservoir solution and buffer containing 100 mM Tris, pH 7.8, 10% (v/v) DMSO, and 8% (w/v) PEG 8000 mixed in a 3:2, 1:2, and 2:3 manner, respectively. Subsequently, the crystal was equilibrated in a droplet of sole buffer pH 7.8 overnight to guarantee a homogeneous pH over the whole crystal. As a cryoprotectant for the apo crystal, 20% (v/v) PEG 400 was used. Data Collection. For data collection, the soaked crystals were transferred into a cryoprotectant solution consisting of 50 mM MES, pH 5.5, 300 mM NaCl, 0.5 mM DTT, 2% (v/v) DMSO, 4% (w/v) PEG 8000, and 30% (v/v) glycerol for 20 s, followed by immediately flash-freezing in liquid nitrogen. Data sets TGT·1 and TGT·2 were collected at the BESSY II (Helmholtz-Zentrum, Berlin, Germany) beamline 14.2 at a wavelength of λ = 0.91841 Å using a Rayonix MX225 CCD detector. Complex structures for TGT Asp102Asn·2 and Asp156Asn·2 were collected at PETRA III (EMBL, Hamburg, Germany) beamline P14 at a wavelength of λ = 1.23953 Å, TGT-apo pH 7.8 at a wavelength of λ = 0.97627 Å using a PILATUS 6M-F detector. To minimize radiation damage, all data sets have been collected at cryoconditions (100 K). All TGT crystals showed the monoclinic space group C2 containing one monomer in the asymmetric unit. Data sets TGT·1 and TGT·2 were processed and scaled with the HKL2000 package.44 Data processing and scaling for TGT Asp102Asn·2, Asp156Asn·2, and TGT-apo pH 7.8 were performed with XDS and XSCALE, respectively.45 Cell dimensions, data collection, and processing statistics are given in Table 3. Structure Determination and Refinement. The coordinates of the TGT apo structure 1PUD served as a starting model for molecular replacement using the program Phaser MR of the ccp4 program suite.46 Structures were refined using the program Phenix (TGT·1, TGT·2, TGT-apo pH 7.8, phenix.refine 1.8.4_1496; Asp102Asn·2, Asp156Asn·2, phenix.refine 1.8.1_1168)47 starting with a first cycle of simulated annealing using default parameters. Further refinement cycles comprised the coordinate xyz, occupancy, and individual Bfactor refinement as well as applying metal restraints for the zinc ion. In case of structures TGT·2 and TGT-apo pH 7.8, the weights between X-ray target and stereochemistry restraints were optimized in addition to individual atomic displacement parameter (ADP) weights refined for TGT·1, TGT·2, and TGT-apo pH 7.8. The temperature

factors of structures TGT·1, TGT·2, and TGT-apo pH 7.8 were refined anisotropically, while for structures TGT Asp102Asn·2 and Asp156Asn·2 a TLS refinement was performed after selecting appropriate TLS groups with the phenix.find_tls_groups option.48 The calculation of the Rfree value comprised a 3−5% fraction of the data. For all structures, amino acid side chains were fitted according to their σA-weighted 2|Fo| − |Fc| and |Fo| − |Fc| electron density obtained in the program Coot.49 After the initial refinement cycles, the zinc ion as well as water and glycerol molecules were implemented in the model. For adding water molecules, the option “update waters” included in Phenix was used after increasing the hydrogen bond length threshold for the solvent−model and solvent−solvent contacts to 2.3 Å. The inserted molecules were visually reviewed afterward. Ligand restraints were generated by the CSD based gradeWebServer [http:// grade.globalphasing.org]. Multiple protein residue conformations were assigned in case a reasonable electron density was observed and were kept during refinement if the side chain with the lowest occupancy showed a value of at least 20%. In case of 1, the occupancy of the morpholinoethyl substituent was refined due to elevated B-factors observed for this portion. Ramachandran plots have been generated with the program PROCHECK.50 For the analysis of temperature factors the program Moleman has been used.51 The burial of molecular ligand portions in the protein binding pocket has been computed using the program MS.52



ASSOCIATED CONTENT

S Supporting Information *

Tables containing measured pKa values for the scaffolds of linbenzoguanines 1, 2, and 4 and lin-benzohypoxanthine 3 as well as calculated pKa values of the amino acid residues within a radius of 12 Å around Cγ of Tyr106 and 1 before and after complex formation. This material is available free of charge via the Internet at http://pubs.acs.org. Accession Codes

Coordinate files were deposited in the PDB with the following access ID: 4PUJ, 4PUK, 4PUL, 4PUM, 4PUN.



AUTHOR INFORMATION

Corresponding Author

*Phone: (+49)6421-28-21313. Fax: (+49)6421-28-28994. Email: klebe@staff.uni-marburg.de. Notes

The authors declare no competing financial interest.



ACKNOWLEDGMENTS M.N. was supported by a grant from the Deutsche Forschungsgesellschaft (KL1204/13-1). The Microcal iTC200 microcalorimeter system (GE Healthcare) used in our studies was financed with kind support of ERC grant no. 268145DrugProfilBind. We are grateful to the support from the beamline staff at BESSY II in Berlin and a travel grant from the Helmholtz-Zentrum für Materialien und Energie in Berlin. We acknowledge the EMBL in Hamburg for their support at MX beamline P14. Work at ETH was supported by the ETH Research Council. We thank Dr. Simone Hörtner for the preparation of compound 4.



ABBREVIATIONS USED DD/AA, donor−donor/acceptor−acceptor motif; H-bond, hydrogen bond; Hepes, 2-(4-(2-hydroxyethyl)-1-piperazinyl)ethanesulfonic acid; IPTG, isopropyl-β-D-thiogalactopyranoside; ITC, isothermal titration calorimetry; J, joule; Kd, dissociation constant; PEOE, partial equalization of orbital J

dx.doi.org/10.1021/jm500401x | J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry

Article

in complex with novel and potent inhibitors unravel pronounced induced-fit adaptations and suggest dimer formation upon substrate binding. J. Mol. Biol. 2007, 370, 492−511. (20) Barandun, L. J.; Immekus, F.; Kohler, P. C.; Tonazzi, S.; Wagner, B.; Wendelspiess, S.; Ritschel, T.; Heine, A.; Kansy, M.; Klebe, G.; Diederich, F. From lin-benzoguanines to lin-benzohypoxanthines as ligands for Zymomonas mobilis tRNA-guanine transglycosylase: replacement of protein-ligand hydrogen bonding by importing water clusters. Chem.Eur. J. 2012, 18, 9246−9257. (21) Durand, J. M. B.; Björk, G. R.; Kuwae, A.; Yoshikawa, M.; Sasakawa, C. The modified nucleoside 2-methylthio-N6-isopentenyladenosine in tRNA of Shigella flexneri is required for expression of virulence genes. J. Bacteriol. 1997, 179, 5777−5782. (22) Grädler, U.; Gerber, H.-D.; Goodenough-Lashua, D. M.; Garcia, G. A.; Ficner, R.; Reuter, K.; Stubbs, M. T.; Klebe, G. A new target for shigellosis: rational design and crystallographic studies of inhibitors of tRNA-guanine transglycosylase. J. Mol. Biol. 2001, 306, 455−467. (23) Kohler, P. C.; Ritschel, T.; Schweizer, W. B.; Klebe, G.; Diederich, F. High-affinity inhibitors of tRNA-guanine transglycosylase replacing the function of a structural water cluster. Chem.Eur. J. 2009, 15, 10809−10817. (24) Hörtner, S. R.; Ritschel, T.; Stengl, B.; Kramer, C.; Schweizer, W. B.; Wagner, B.; Kansy, M.; Klebe, G.; Diederich, F. Potent inhibitors of tRNA-guanine transglycosylase, an enzyme linked to the pathogenicity of the Shigella bacterium: charge-assisted hydrogen bonding. Angew. Chem., Int. Ed. 2007, 46, 8266−8269; Angew. Chem. 2007, 119, 8414−8417. (25) Brenk, R.; Stubbs, M. T.; Heine, A.; Reuter, K.; Klebe, G. Flexible adaptations in the structure of the tRNA-modifying enzyme tRNA-guanine transglycosylase and their implications for substrate selectivity, reaction mechanism and structure-based drug design. ChemBioChem 2003, 4, 1066−1077. (26) Ritschel, T.; Hoertner, S.; Heine, A.; Diederich, F.; Klebe, G. Crystal structure analysis and in silico pKa calculations suggest strong pKa shifts of ligands as driving force for high-affinity binding to TGT. ChemBioChem 2009, 10, 716−727. (27) Fukada, H.; Takahashi, K. Enthalpy and heat capacity changes for the proton dissociation of various buffer components in 0.1 M potassium chloride. Proteins 1998, 33, 159−166. (28) Christensen, J. J.; Hansen, L. D.; Izatt, R. M. Handbook of Proton Ionization Heats and Related Thermodynamic Quantities; WileyInterscience: New York, 1976. (29) Hall, H. K. Correlation of the base strengths of amines. J. Am. Chem. Soc. 1957, 79, 5441−5444. (30) Czodrowski, P.; Dramburg, I.; Sotriffer, C. A.; Klebe, G. Development, validation, and application of adapted PEOE charges to estimate pKa values of functional groups in protein-ligand complexes. Proteins 2006, 65, 424−437. (31) Romier, C.; Reuter, K.; Suck, D.; Ficner, R. Mutagenesis and crystallographic studies of Zymomonas mobilis tRNA-guanine transglycosylase reveal aspartate 102 as the active site nucleophile. Biochemistry 1996, 35, 15734−15739. (32) Tidten, N.; Stengl, B.; Heine, A.; Garcia, G. A.; Klebe, G.; Reuter, K. Glutamate versus glutamine exchange swaps substrate selectivity in tRNA-guanine transglycosylase: insight into the regulation of substrate selectivity by kinetic and crystallographic studies. J. Mol. Biol. 2007, 374, 764−776. (33) Romier, C.; Reuter, K.; Suck, D.; Ficner, R. Crystal structure of tRNA-guanine transglycosylase: RNA modification by base exchange. EMBO J. 1996, 15, 2850−2857. (34) Jorgensen, W. L.; Pranata, J. Importance of secondary interactions in triply hydrogen-bonded complexesguanine−cytosine vs uracil-2,6-diaminopyridine. J. Am. Chem. Soc. 1990, 112, 2008− 2010. (35) Murray, T. J.; Zimmerman, S. C. New triply hydrogen-bonded complexes with highly variable stabilities. J. Am. Chem. Soc. 1992, 114, 4010−4011. (36) Bogan, A. A.; Thorn, K. S. Anatomy of hot spots in protein interfaces. J. Mol. Biol. 1998, 280, 1−9.

electronegativities; TGT, tRNA-guanine transglycosylase; Tricine, N-(tris(hydroxymethyl)methyl)glycine; Z. mobilis, Zymomonas mobiliz



REFERENCES

(1) Burnouf, D.; Ennifar, E.; Guedich, S.; Puffer, B.; Hoffmann, G.; Bec, G.; Disdier, F.; Baltzinger, M.; Dumas, P. kinITC: A new method for obtaining joint thermodynamic and kinetic data by isothermal titration calorimetry. J. Am. Chem. Soc. 2012, 134, 559−565. (2) Tsamaloukas, A. D.; Keller, S.; Heerklotz, H. Uptake and release protocol for assessing membrane binding and permeation by way of isothermal titration calorimetry. Nature Protoc. 2007, 2, 695−704. (3) Zhang, Y.-L.; Zhang, Z.-Y. Low-affinity binding determined by titration calorimetry using a high-affinity coupling ligand: a thermodynamic study of ligand binding to protein tyrosine phosphatase 1B. Anal. Biochem. 1998, 261, 139−148. (4) Velazquez-Campoy, A.; Freire, E. Isothermal titration calorimetry to determine association constants for high-affinity ligands. Nature Protoc. 2006, 1, 186−191. (5) Ladbury, J. E.; Klebe, G.; Freire, E. Adding calorimetric data to decision making in lead discovery: a hot tip. Nature Rev. Drug Discovery 2010, 9, 23−27. (6) Baker, B. M.; Murphy, K. P. Evaluation of linked protonation effects in protein binding reactions using isothermal titration calorimetry. Biophys. J. 1996, 71, 2049−2055. (7) Raffa, R. B.; Stagliano, G. W.; Spencer, S. D. Protonation effect on drug affinity. Eur. J. Pharmacol. 2004, 483, 323−324. (8) Baum, B.; Muley, L.; Heine, A.; Smolinski, M.; Hangauer, D.; Klebe, G. Think twice: understanding the high potency of bis(phenyl)methane inhibitors of thrombin. J. Mol. Biol. 2009, 391, 552−564. (9) Steuber, H.; Czodrowski, P.; Sotriffer, C. A.; Klebe, G. Tracing changes in protonation: a prerequisite to factorize thermodynamic data of inhibitor binding to aldose reductase. J. Mol. Biol. 2007, 373, 1305− 1320. (10) Czodrowski, P.; Sotriffer, C. A.; Klebe, G. Protonation changes upon ligand binding to trypsin and thrombin: structural interpretation based on pKa calculations and ITC experiments. J. Mol. Biol. 2007, 367, 1347−1356. (11) Dullweber, F.; Stubbs, M. T.; Musil, Đ.; Stürzebecher, J.; Klebe, G. Factorising ligand affinity: a combined thermodynamic and crystallographic study of trypsin and thrombin inhibition. J. Mol. Biol. 2001, 313, 593−614. (12) Biela, A.; Khayat, M.; Tan, H.; Kong, J.; Heine, A.; Hangauer, D.; Klebe, G. Impact of ligand and protein desolvation on ligand binding to the S1 pocket of thrombin. J. Mol. Biol. 2012, 418, 350− 366. (13) Biela, A.; Sielaff, F.; Terwesten, F.; Heine, A.; Steinmetzer, T.; Klebe, G. Ligand binding stepwise disrupts water network in thrombin: enthalpic and entropic changes reveal classical hydrophobic effect. J. Med. Chem. 2012, 55, 6094−6110. (14) Testa, B.; Carrupt, P.-A.; Gaillard, P.; Billois, F.; Weber, P. Lipophilicity in molecular modeling. Pharm. Res. 1996, 13, 335−343. (15) El Tayar, N.; Tsai, R.-S.; Testa, B.; Carrupt, P.-A.; Leo, A. Partitioning of solutes in different solvent systems: the contribution of hydrogen-bonding capacity and polarity. J. Pharm. Sci. 1991, 80, 590− 598. (16) Conradi, R. A.; Hilgers, A. R.; Ho, N. F. H.; Burton, P. S. The influence of peptide structure on transport across Caco-2 cells. Pharm. Res. 1991, 8, 1453−1460. (17) Spencer, J. N.; Gleim, J. E.; Blevins, C. H.; Garrett, R. C.; Mayer, F. J. Enthalpies of solution and transfer enthalpies. An analysis of the pure base calorimetric method for the determination of hydrogenbond enthalpies. J. Phys. Chem. 1979, 83, 1249−1255. (18) Schanker, L. S. Passage of drugs across body membranes. Pharmacol. Rev. 1962, 14, 501−530. (19) Stengl, B.; Meyer, E. A.; Heine, A.; Brenk, R.; Diederich, F.; Klebe, G. Crystal structures of tRNA-guanine transglycosylase (TGT) K

dx.doi.org/10.1021/jm500401x | J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry

Article

(37) Mizoue, L. S.; Tellinghuisen, J. The role of backlash in the “first injection anomaly” in isothermal titration calorimetry. Anal. Biochem. 2004, 326, 125−127. (38) Reuter, K.; Ficner, R. Sequence analysis and overexpression of the Zymomonas mobilis tgt gene encoding tRNA-guanine transglycosylase: purification and biochemical characterization of the enzyme. J. Bacteriol. 1995, 177, 5284−5288. (39) Romier, C.; Ficner, R.; Reuter, K.; Suck, D. Purification, crystallization, and preliminary X-ray diffraction studies of tRNAguanine transglycosylase from Zymomonas mobilis. Proteins 1996, 24, 516−519. (40) Laemmli, U. K. Cleavage of structural proteins during the assembly of the head of bacteriophage T4. Nature 1970, 227, 680− 685. (41) Barken, F. M.; Gasteiger, E. L. Excitability of a penicillin-induced cortical epileptic focus. Exp. Neurol. 1980, 70, 539−547. (42) Word, J. M.; Lovell, S. C.; Richardson, J. S.; Richardson, D. C. Asparagine and glutamine: using hydrogen atom contacts in the choice of side-chain amide orientation. J. Mol. Biol. 1999, 285, 1735−1747. (43) Bashford, D. An object-oriented programming suite for electrostatic effects in biological molecules. An experience report on the MEAD project. In Scientific Computing in Object-Oriented Parallel Environments; Ishikawa, Y., Oldehoeft, R., Reynders, J., Tholburn, M., Eds.; Springer: Berlin/Heidelberg, 1997; Vol. 1343, pp 233−240. (44) Otwinowski, Z.; Minor, W. Processing of X-ray diffraction data collected in oscillation mode. Methods Enzymol. 1997, 276, 307−326. (45) Kabsch, W. XDS. Acta Crystallogr., Sect. D: Biol. Crystallogr. 2010, 66, 125−132. (46) McCoy, A. J. Solving structures of protein complexes by molecular replacement with Phaser. Acta Crystallogr., Sect. D: Biol. Crystallogr. 2007, 63, 32−41. (47) Adams, P. D.; Afonine, P. V.; Bunkóczi, G.; Chen, V. B.; Davis, I. W.; Echols, N.; Headd, J. J.; Hung, L.-W.; Kapral, G. J.; GrosseKunstleve, R. W.; McCoy, A. J.; Moriarty, N. W.; Oeffner, R.; Read, R. J.; Richardson, D. C.; Richardson, J. S.; Terwilliger, T. C.; Zwart, P. H. PHENIX: a comprehensive Python-based system for macromolecular structure solution. Acta Crystallogr., Sect. D: Biol. Crystallogr. 2010, 66, 213−221. (48) Painter, J.; Merritt, E. A. Optimal description of a protein structure in terms of multiple groups undergoing TLS motion. Acta Crystallogr., Sect. D: Biol. Crystallogr. 2006, 62, 439−450. (49) Emsley, P.; Cowtan, K. Coot: model-building tools for molecular graphics. Acta Crystallogr., Sect. D: Biol. Crystallogr. 2004, 60, 2126− 2132. (50) Laskowski, R. A.; Mac Arthur, M. W.; Moss, D. S.; Thornton, J. M. PROCHECKa program to check the stereochemical quality of protein structures. J. Appl. Crystallogr. 1993, 26, 283−291. (51) Kleywegt, G. J.; Zou, J. Y.; Kjeldgaard, M.; Jones, T. A. Around, O. In International Tables for Crystallography; Rossmann, M. G., Arnold, E., Eds.; Kluwer Academic Publishers: Dordrecht, 2001; Vol. F, pp 353−356. (52) Connolly, M. L. Analytical molecular surface calculation. J. Appl. Crystallogr. 1983, 16, 548−558.

L

dx.doi.org/10.1021/jm500401x | J. Med. Chem. XXXX, XXX, XXX−XXX